Upload
builien
View
243
Download
6
Embed Size (px)
Citation preview
1
Running head: Emotional ornamentation
Emotional ornamentation in performances of a Handel sonata
Renee Timmers and Richard Ashley
School of Music, Northwestern University
2
Abstract
Ornamentation is one aspect of music associated with emotional affect in Baroque
music. In an empirical study, the relationship between ornamentation and emotion
was investigated by asking a violinist and flutist to ornament three melodies in
different ways to express four emotions: happiness, sadness, love and anger. The
performers adapted the type of ornaments to the instructed emotion as well as the
characteristics of the ornaments. The flutist specifically varied the duration, timing
and complexity of the ornamentation, while the violinist varied the complexity,
density and sound level of the performances. The ability of the performers to
communicate the emotions was tested in a listening experiment. Communication was
found to be generally successful, with the exception of the communication of
happiness. This success was not due to general consensus about the expression of
emotions through ornamentation. Rather the listeners were sensitive to a performer’s
specific use of ornamentation.
Emotion, ornamentation, affects in baroque music, performance, perception
3
Introduction
The aim of the present study was to investigate the relationship between variations
in melodic ornamentation and perceived emotion in the context of modern
performances of a Handel sonata. The research aimed to contribute to the
understanding of the attribution of emotional qualities to music and the
communication of emotion using music. The questions asked were: What musical
qualities relate to perceived emotions, and, specifically, what is the role, if any, of
ornamentation in the communication of emotion for modern-day musicians? These
questions are distinct from the question of how listeners come to experience emotions
in response to music. Listeners’ experience of emotions may be more subjective and
socially dependent than their perception of emotion (see for discussions Scherer &
Zentner, 2001; Scherer, 2004; Gabrielsson, 2001/2002).
Previous research has shown that listeners are quite consistent in labeling music
with an emotional quality, especially where it concerns basic emotions (Gabrielsson
& Juslin, 2003). This research has started to distinguish relevant musical features that
are responsible for labeling music with a specific emotional category. Perhaps not
surprisingly, almost every musical feature, including aspects of the performance, may
influence the perception of emotions; this may include harmony, rhythm, melody,
tempo, dynamics, timbre, vibrato, and articulation (see e.g. Gabrielsson & Juslin,
1996; 2003; Hevner, 1935; Sundberg, Iwarsson & Hagegård, 1995). Some of these
features are apparently more important than others. For example, tempo and mode are
both salient features, but, when contrasted, tempo can overrule the effect of mode to
distinguish happy from sad sounding music (Gagnon & Peretz, 2003).
The exact influence of musical features on the perception of emotions is, however,
not always clear and consistent. In their overview of musical variables and their
4
emotional connotations, Gabrielsson and Lindström (2001) noted that musical
features are often associated with many (even contrasting) emotions, and interactions
between features are often found. For example, the effects of melodic contour and
rhythm interact when judging melodies as sounding happy or sad.
To explain relationships between emotion and music, at least two theories have
been offered: 1) a theory based on musical cues and 2) a theory based on extended
musical structure (for summaries, see Gabrielsson & Juslin, 2003 and Sloboda &
Juslin, 2001). The theory that associates perception of emotion with musical cues
emphasizes the communication of specific emotions, such as happiness and sadness,
based on local and global characteristics of the music. Cues may be either acoustic or
structural. Acoustic cues include speed, loudness, pitch, timbre, attack, and decay of
events, while structural cues include tonal mode (major or minor), melodic contour,
and harmonic and melodic intervals.
Cues may have an origin in vocal expressions of emotions in speech (Juslin &
Laukka, 2003) and in infant directed non-verbal communication (‘motherese’, e.g.
Trehub & Nakata, 2001/2002). Cues may also have a biological basis, which is the
case for dissonance, which is innately disliked (Trainor & Heinmiller, 1998). In
addition, cultural cues are established within a tradition and include, for example, the
different meanings of scales (see Balkwill & Thompson, 1999 for a clear
differentiation between cultural and psychoacoustic cues). Cultural cues are
associational, rather than iconic (a distinction made by Sloboda & Juslin, 2001),
meaning that cultural cues and emotions are related by convention rather than by
similarity.
The other theory concerned with musical structure is related to the theories of
emotion of Meyer (1956) and Langer (1942). It emphasizes the temporal structure of
5
music, i.e. how music unfolds over time, and its analogy with the temporal structure
of emotions. To give a very short outline, according to Meyer, music may arouse
emotions by setting up expectations that are rewarded, violated, or suspended. This
parallels the cognitive framework for emotions when an individual’s goals are
frustrated or rewarded (Mandler, 1975; Oatley, 1992). Very briefly, in Langer’s
theory, patterns of ebb and flow in music evoke emotions by analogy to general forms
of feeling. No specific emotional categories are communicated, but rather patterns of
tension and release.
A few empirical studies have provided evidence for these theories. Krumhansl
(1996) and Lerdahl and Krumhansl (2007) related perceived tension in music to
points of tension and relaxation, mostly harmonic in nature, in line with Meyer’s
theory. More generally, the fluctuations in tension and emotionality as recorded with
continuous response methods suggest the importance of temporal structure for the
affective impact of musical events (see e.g. Sloboda & Lehmann, 2001; Krumhansl,
1998; Schubert, 2001).
These two types of theories are rather different from each other and they may well
tap into two distinct processes. While the first is concerned with the recognition of an
emotional character based on acoustic or cultural cues, the second concerns emotional
tension build-up and release over time. Another difference is that the cue-based theory
does not consider structural context. Recognition of an emotion depends on an
addition of cues; which emotion is recognized depends on which emotion gets most
evidence, as in Juslin’s adaptation of the Brunswikian lens model (Juslin, 2000). In
this summation, the kind of cues may partly depend on participants’ (cultural)
background, but the interpretation of cues is fixed within a certain context. For
example, loud sounds will generally increase the evidence for anger and decrease the
6
evidence for sadness. The interpretation of cues is established within a stable
framework of emotional connotations of sounds and musical elements. This contrasts
with Meyer’s account that makes the meaning of events primarily relative to structural
context and relationships with other events (embodied musical meaning in Meyer’s
terms, 1956).
The issue central to the current study concerned the possible role of ornamentation
in the communication of emotions in music. Ornaments are notes that can be added to
a melody as a kind of melodic variation. Ornaments are often tied to a melody note to
which they belong and which they elaborate. For Western classical music,
ornamentation was an important part of performance practice within the Baroque era.
Several treatises from that time contain elaborate sections on ornamentation and
describe what performers should and should not do (for summaries see Donington,
1963 and Wessel, 1979). In some of these, a relation between ornamentation and
emotion is suggested. For example, Carl P. E. Bach (1752) wrote that for each affect,
be it happy or sad, ‘ornaments will lend a fitting assistance’. And Mattheson (1739)
observed that ‘it is necessary that [a singer] knows how to perform a precomposed
melody with much grace, ornament and artistry: [This] is reading with expression and
good style’ (p. 265). Mattheson (1739) included ornaments among the means that
were used to emphasize words or notes. And the aim of ‘emphasis’ is to point ‘toward
the emotion, and illuminate the sense or meaning of the performance’ (p. 370).
These treatises provide some suggestions that specific ornaments may express one
affect and not another, but clear predictions are difficult to draw from this literature.
Therefore our predictions for the relationships between ornaments and emotions are
based on the emotion theories summarized above. First of all, we expected that
ornaments function as cultural cues, which means that they are a stylistic convention:
7
Specific types of ornaments, such as an appoggiatura or trill, are associated with
specific emotions, such as sadness or happiness. Secondly, ornamentation is a way of
varying the local characteristics of a melodic line. For example, variation of the
number of notes or the duration of the ornament varies local tempo, while variation of
the pitch of the ornament may make a passage more dissonant. In this way, ornaments
function as acoustic cues. Finally, ornamentation is a melodic variation at a certain
structural position. The ornament adds an aspect of surprise and unexpectedness. The
structural position of this variation is one factor that influences the strength of
surprise. For example, violations of expectancies may be more salient for strong
metrical positions and for structural positions later in a phrase. In this way, the
structural position of an ornament may influence its meaning.
We report two studies that explored these three ways in which ornaments may
influence perception of emotion in modern-day practice. In the first study, two
performers, experienced with ornamenting Baroque music, were asked to ornament
four musical fragments taken from a Handel sonata to express different emotions:
mild and intense happiness, mild and intense love, mild and intense anger, and mild
and intense sadness. Love differs from the other emotions, because it is not generally
considered to be a basic emotion category. However, it is an important emotion in
theories of Clynes (1977) and Ortony, Clore & Collins (1988) and important in many
musical contexts. Furthermore, these emotions were chosen to vary valence, activity
and intensity systematically.
According to the circumplex model of emotions (Russell, 1980), emotions differ
along the two dimensions of valence and activity. Valence distinguishes negative
emotions, such as sadness and anger, from positive emotions, such as happiness and
love. Activity distinguishes emotions with higher levels of arousal, such as anger and
8
happiness, from emotions with lower levels of arousal, such as sadness and lovingly
(which was the term used in the instructions for the participants). Additionally,
emotions may vary in intensity (e.g. Ortony, Clore, Collins, 1988). The intensity of an
emotion may distinguish cold anger from hot anger, or mild from intense sadness.
Several studies have suggested the usefulness of these dimensions for the
interpretation of emotions perceived in music (e.g. Ilie & Thompson, 2006; Juslin &
Laukka, 2001, Schubert, 2001; Sloboda & Lehmann, 2001). Moreover, happiness,
sadness, and love are emotions that are often mentioned in music treatises, when
examples of affects are given (for a list of common affects and terminologies, see
Wessel, 1979).
In our study, we asked the performers to ornament musical fragments with a variety
of emotional expressions. This procedure is taken from speech research in which
actors are asked to read a text aloud in different ways (see e.g. Kramer, 1964; Scherer,
Banse, Wallbott & Goldbeck, 1991). The mode of expression is not related to the
meaning of the text. Instead a factual text without emotional meaning is often used to
easily separate the influence of the actors’ reading from the influence of the meaning
of the text. In music research, this procedure is also used, asking musicians to perform
melodies with different emotional expressions (for an overview, see Juslin & Laukka,
2003). Although it may be possible that melodies are emotionally “neutral”, it is more
likely that most melodies have some emotional connotation that influence the
perception of emotions. We took this possible effect of musical fragment into account
and separated it from the effect of performance.
The two performers ornamented the music according to their own ideas and
intuition. Despite this freedom, we expected that the applied ornamentation would fall
within the conventions of Baroque music and can be classified according to the
9
categories distinguished by performance practice treatises of that time, such as
appoggiaturas, trills and turns. More specifically, in the analysis of the ornamentation,
we first classified the ornaments according to categories distinguished by Bach (1753)
and then analyzed the characteristics of the ornaments, such as the melodic and
harmonic relationships of the ornament to the main note and chord, again following
Bach’s distinctions. The structural location of the ornament included its serial position
and metrical position within the musical fragment. Finally, we examined the relations
between these analyzed characteristics of the ornamentation and the instructed
emotions of the performances.
In the second study, listeners’ perception of emotions in the recorded performances
was examined. Musically trained participants (not specialists in Baroque music)
indicated perceived emotions by rating the presence of the four emotions on separate
unipolar scales. We then related their ratings to the intended emotions and to the
analyzed characteristics of the ornamentation.
In this way, the performers’ strategies to express emotions could be distinguished
from listeners’ attribution of emotional qualities to music. Mode and tempo were
fixed by the musical fragment that was ornamented and by asking the performers to
play along with a pre-recorded accompaniment. This excluded these relatively salient
cues to perceived emotion and allowed us to focus on the effects of ornamentation.
However, other cues such as dynamics, timbre, and articulation could still be varied.
Average sound level was therefore analyzed as an additional potential acoustic cue for
emotions. The other additional cues were not considered in this study due to the
complexity of the acoustic analyses that would have been required.
To summarize, the questions of this study were: 1) Do performers instructed to
convey different emotions via ornamentation adopt consistent strategies in doing this?
10
2) Do performers and listeners with similar backgrounds share knowledge about
communication of emotion through ornamentation? 3) How does this knowledge fit
the hypothesized relationships between perceived emotion and type, characteristics, or
position of ornamentation?
Study 1: Performers’ ornamentation of melodies
Method
Materials. Materials for the performances were the opening bars of the first (bars 1-
5), second (bars 1-16), third (bars 1-7) and fourth movement (bars 1-13) of Handel’s
sonata HWV360 in G minor for recorder and basso continuo (Op. 1, No. 2). This is an
early sonata of Handel, composed between 1724 and 1726. The approximate tempo
was 78 eighth notes per minute for fragment 1 (abbreviation F1), 96 quarter notes per
minute for fragment 2 (abbreviation F2), 58 half notes per minute for fragment 3
(abbreviation F3), and 112 quarter notes per minute for fragment 4 (abbreviation F4).
Scores of fragments 1, 3 and 4 are given in Figure 1. F2 is not shown, since it was
excluded from the analysis (see below). The last chord of F3 is omitted in the figure
to save an extra line. Complete scores were given to the performers.
Insert Figure 1 around here
Additional material for exemplification of the intended emotions to the performers
was collected from two CDs with performances of Handel’s solo sonatas.[footnote 1]
These performances of a variety of sonatas and our recordings of the piano
accompaniment of sonata HWV360 were put on CD. Track information is listed in
Appendix 1.
11
Performers. A female flutist and a male violinist performed the solo part of the
fragments from the Handel sonata. They are both professional musicians who perform
regularly on instruments of the Baroque period as well as on modern instruments. A
highly skilled pianist from the Northwestern University School of Music performed
the basso continuo of the music.
Procedure. Three recording sessions were organized in a 450 seat concert hall of
Northwestern University. In the first recording session, the accompaniment of the
Handel sonata was recorded. The pianist was instructed to perform the basso continuo
of the entire sonata with neutral expression, approximately at an indicated tempo,
though with small, musically appropriate variations in tempo.
In the second and third sessions, the ornamented melodies were recorded. Three
weeks before these recordings, scores, written instructions, and a CD with audio
tracks (Appendix 1) were sent to the performers with the following instruction:
“Please prepare the beginnings of each fragment [bars and tempi for each fragment
were indicated] first of all without any additions of ornaments or changes of the score
and secondly, with different ornamentation in order to express the following eight
emotions: Happy mild (gayement) and intense (brillant), loving mild (cantabile) and
intense (con brio), sad mild (doloroso) and intense (lagrimoso), and angry mild
(ardito) and intense (furioso). The choice of ornamentation is left to your taste and
interpretation”. In addition, a practice procedure was suggested that included four
steps: (1) choosing a mood and fragment to perform, (2) listening to the track that
exemplifies an expression of the emotion, (3) imagining the performance of the
Handel sonata, and (4) performing the fragment along with the piano accompaniment.
The Italian and French indications of the moods and sounding examples of the
moods were included in addition to emotion words to help make the instructions more
12
natural for musicians. Sounding examples eliminated potential doubt about the
possibility of expressing the instructed moods in music. The examples did not contain
much ornamentation and therefore should not have influenced the performances in
this respect.
Separate recording sessions were held for the flutist and the violinist. At the
recording session, the respective soloist performed each version of each fragment as
often as needed to achieve a satisfactory performance. In most instances, the first
performance was sufficiently satisfying. The soloist performed the music along with
the accompaniment, which he/she heard over a headphone that covered only one ear.
Headphones were used in order to be able to record the soloist separately from the
accompaniment.
The order in which the versions were recorded was fixed. The recording session
started with the neutral version of F1, followed by the mild and intense happy version
of this fragment, the mild and intense loving version, the mild and intense sad version,
and the mild and intense angry version. The same order was followed for fragments 2,
3 and 4.
The best performance was selected from the performances of the versions that were
recorded more than once. This was generally the last version, or whichever had best
satisfied the performer. This led to a collection of one performance by the flutist and
one performance by the violinist of the nine conditions of the four fragments, a total
of 72 performances (2 performers * 9 versions * 4 fragments).
Analysis
As a preparation for the analysis of the ornamentation, we first made a transcription
of the performances and separated ornamental notes from structural (notated) notes.
All notes that were present in the transcription but were not present in the original
13
score were defined as ornaments. Secondly, we identified the main note for each
ornamental note. The main note was defined to be the first structural note following
the ornamental note(s).
These steps and the following steps of the analysis were done as closely in
agreement with Baroque performance practice and music theoretical literature as
possible. However, sometimes a systematic approach was preferred above a
historical-musicologically oriented interpretation. For example, the rule that all notes
that do not occur in the score are ornamental and all score-notes are structural could
have had exceptions, when analyzed music-theoretically. Nevertheless, we felt that
the rule suits the current study well, since the score is the starting point for the
performers.
Type of ornament. To classify the ornaments used by the performers, we took the
ornament types distinguished by Bach (1753): the appoggiatura, turn, mordent, trill,
slide, the compound appoggiatura, and the slide. In addition, Bach distinguished the
snap, which is not usually considered as a separate category, but rather as a variant of
the mordent, so we did not include this distinction (Donington, 1963). The ornaments
we considered may be understood as follows (see also Figure 2):
• A single appoggiatura is a one-note ornament that is tied to the following main
note.
• A double appoggiatura consists of two notes that both fall or rise to the following
main note.
• A trill is a rapid alternation of the main note and a note a semitone or whole tone
above the main note. The trill often starts from the upper note and ends on the
main note.
14
• A turn ornaments the main note by playing around it: one note above the main
note, the main note, one note below and a return to the main note.
• A mordent is an alternation of the main note with a neighboring note that starts
with the main note: main note, one note above or below the main note, and a
return to the main note.
• A slide fills a melodic interval by stepwise intervals in one direction. In the
analyses, the slide should have at least three tones going up or down in the same
direction to be recognized as such.
Insert Figure 2 around here
To these classes, we added the substitute and the arpeggio. An ornament is a
substitute when the main note is eliminated and replaced by the ornamental note. An
ornament is an arpeggio when it consists of broken chords or intervals larger than a
whole tone.
Figure 3 gives examples of transcribed performances and labeled ornaments. For
example, the first line gives the first bars of F1 in the “Intense Happy” condition as
performed by the flutist. She introduced arpeggios and a trill in these first two bars.
Arrows indicate the main notes that immediately follow the ornaments. The second
line gives the first bars of F1 for the “Intense Lovingly” condition for the flutist. In
this condition, she introduced a mordent that precedes a slide, which in turn precedes
a slide towards the main note. This ornamentation is repeated in a shortened version
(mordent followed by one slide towards the main note). At the end of bar one, we
distinguish two single appoggiaturas and, in bar 2, we distinguish a trill, a double
appoggiatura and a slide.
15
Insert Figure 3 around here
The classification of these ornaments was not always straightforward, as it
depended on the identification of main notes and on the grouping of notes as one
ornament. We applied three rules as consistently as possible: 1) The main note is the
score note at its original position or delayed and shortened by an ornament. 2) The
notes preceding the main note that are not score notes are ornamental notes. 3)
Ornamental notes are grouped into traditional categories where possible.
Characteristics of the ornaments. For the characterization of each ornament, we also
took into account distinctions made in performance treatises. Bach (1753), Quantz
(1752), and Donington (1963) differentiate between long and short appoggiaturas,
ascending and descending appoggiaturas, appoggiaturas stealing time from the main
note and sparing the main note, and between consonant and dissonant appoggiaturas.
These characterizations of appoggiaturas can also be applied to the other ornaments.
• A long appoggiatura should be at least half the duration of the main note,
otherwise it is short. Compare, for example, the second long with the first short
appoggiatura in the third line of Figure 3.
• In an ascending appoggiatura, the ornament ascends to the following note. In a
descending appoggiatura, the ornament descends to the following note. Compare,
for example, the second ascending with the first descending appoggiatura in the
seventh line of Figure 3.
• An appoggiatura that steals time from the main note shortens and shifts the main
note. This type of timing of the ornament is referred to as “on-main” and is often
“on the beat.” An appoggiatura that spares the main note precedes the main note
16
and leaves the onset timing of the main note unaffected. [footnote2] This type of
timing is referred to as “pre-main”. This manner of timing is often, but not always,
“before the beat.” Compare, for example, the second pre-main with the first on-
main appoggiatura in the seventh line of Figure 3.
• The consonance or dissonance of an appoggiatura is determined by the interval of
the appoggiatura with the root of the chord that is performed at the same time. If
this interval is an octave, prime, perfect fifth, major or minor third or sixth, it is
consonant; otherwise it is dissonant. Compare, for example, the appoggiatura and
substitute in bar 1 of the seventh line of Figure 3, which are consonant tones, with
the two appoggiaturas in bar 2 of this performance, which are dissonant tones.
To these characteristics, we added complexity and density. Complexity
distinguishes simple ornaments that consist of only one note (the single appoggiatura
and the substitute) from compound ornaments that consist of more than one note (the
turn, mordent, slide and double appoggiatura). Density is simply the number of
ornaments added in a musical fragment.
Average sound level. The average sound pressure level per performance in dB was
calculated using PRAAT (Boersma and Weenink, 2006).[footnote 3]
Structural position of the ornaments. Three dependent variables related to structural
position were defined: metrical position, bar position, and position within the
fragment. Metrical position defines the position of an ornamental note within a
metrical hierarchy, while bar and position within the fragment define the serial
position of an ornamental note within a bar or fragment.
Five metrical levels were distinguished: the bar level, the first division of a bar in
two or three, the beat level, the half beat level, and the quarter beat level. The metrical
weight of a specific metrical position was defined as the sum of the levels at that
17
position, which is highest for the first beat of a bar (5) and lowest for the smallest
subdivision (1).
Two bar positions were distinguished: first half or second half of the bar. Three
positions within fragments were distinguished: first third, second third, and last third
of the fragment. Note that these structural characteristics are simple distinctions that
do only partial justice to the theories of Meyer (1956) and Langer (1942). Our tests
for the effect of structural position can therefore only be seen as a first step.
Figure 4 gives an overview of the analysis procedure: It shows the successive steps
of 1) transcription, 2) comparison with the original score to distinguish ornaments
from structural notes, and 3) classification of the ornaments, characterization of the
ornaments, and the definition of the position of the ornaments. It also shows the
summaries per condition that were used in the further analyses of the ornaments: The
number of ornaments of a certain type, with a certain character, and at a certain
position were summed per condition and this number was divided by the total number
of ornaments per condition, which gave the percentage of ornaments of a certain type,
character or position per condition. This made all variables irrespective of the number
of ornaments. The total number of ornaments was included as a separate variable
(“density”).
Insert Figure 4 around here
To examine the relationship between emotion and ornamentation, the effect of
fragment on the use of ornaments needed to be distinguished from the effect of
emotion instructions. The fragments are rather different from each other and influence
the ornamentation that can be applied. For example, F4 is relatively fast and therefore
18
contains relatively many single appoggiaturas and fewer compound appoggiaturas
such as turns and slides. To eliminate such effects of fragment, the average percentage
per fragment was calculated for ornaments of a certain type, characteristic, and at a
certain position. This average was subtracted from the average per performance to
obtain percentage deviations that were then analyzed further.
To determine the association between the instruction of an emotion (coded as 1 in
the target performances and 0 in the other performances) and the presence of an
ornament in performance, correlations were calculated between the coded instructed
emotions and the percentage deviations for ornaments. These correlations were based
on 24 data points: 3 fragments x 8 instructions per performer. A similar procedure
was used to calculate correlations with the activity of the instructed emotion (coded as
0 or 1) and with the valence of the instructed emotion (also coded as 0 and 1). A
separate analysis was conducted to determine the effect of emotional intensity.
The neutral version of all fragments and all performances of F2 were excluded from
analysis. The tempo of the F2 accompaniment turned out to be too slow to be
comfortable for the performers, so that there were difficulties of synchronization
between soloist and accompaniment.
Results
Flute: type of ornamentation. The top rows of Table 1 list the significant correlations
with the type of ornament for the flute performances. As can be seen, happy
performances showed a positive correlation with the presence of turns, while loving
performances showed a positive correlation with the presence of slides. Sad
performances correlated negatively with the presence of trills and arpeggios, and
positively with the presence of single appoggiaturas, while angry performances
correlated negatively with the presence of single appoggiaturas, and positively with
19
arpeggios. High activity performances showed more trills, arpeggios and fewer single
appoggiaturas than low activity performances, while performances with positive
valence showed more slides than performances with negative valence.
Insert Table 1 around here
Flute: characteristics of ornamentation. The middle rows of Table 1 show the
significant correlations between instructed emotions and characteristics of the
ornamentation. These characteristics included complexity, density, timing, duration,
direction, and harmony. Each of these characteristics correlates significantly with at
least one instructed emotion. Sad performances had, in contrast to happy
performances, relatively many ornaments occurring at the original time of the main
note (on-main). Sad performances had relatively many long ornaments, fewer
complex ornaments, and more dissonant harmony than the other performances. Angry
performances had fewer long ornaments, more complex ornaments, and more
downward ornaments. Activity of emotions was negatively associated with duration
and positively associated with the complexity of ornaments. Positive compared to
negative valence of emotions was associated with pre-main timing rather than on-
main timing. This indicates an association between the activity of emotions and
duration and between the valence of emotions and timing.
Flute: Sound level of performance. There were no significant correlations between
instructed emotion and sound level of the performances, indicating that sound level
was not a deliberately manipulated parameter for the expression of intended emotions.
Flute: Structural position of ornaments. The only significant correlations between
instructed emotion and structural position were with metrical position. As can be seen
20
in Table 1, bottom row, happy ornaments showed a negative correlation with metrical
weight and sad ornaments showed a positive correlation. Both valence and activity of
instructed emotion were negatively associated with metrical weight: More ornaments
fell on weak beats in conditions with positive and active emotions. These
relationships may have different origins: The ornaments at weak beats in positive
emotions relate to the ornaments timed relatively often pre-main, given that the main
note was often timed on the beat, while the ornaments at weak beats in active
emotions relate to the greater complexity of ornaments in performances with active
emotions.
Violin: Type of ornamentation. Significant correlations for the violin performances
between instructed emotions and type of ornaments are reported in the top half of
Table 2. There was a positive correlation between happy performances and the
presence of trills, and between loving performances and the presence of single
appoggiaturas. Sad performances were negatively associated with double
appoggiaturas and positively associated with substitutes. Angry performances were
positively associated with the presence of double appoggiaturas.
Activity of emotions was negatively associated with the presence of single
appoggiaturas and positively associated with the presence of double appoggiaturas,
while positive valence of emotions was associated with the presence of trills. Thus,
the violinist’s use of trills was different from the flutist, for whom the use of trills was
associated with high activity emotional performances.
Insert Table 2 around here
21
Violin: Characteristics of ornamentation. The middle rows of Table 2 show the
significant correlations between instructed emotions and characteristics of the
ornaments. The only significant relationships are with sad and angry performances,
which are the most passive and the most active emotion, respectively. Sad
performances were negatively associated with complexity and density, while angry
and high activity performances were positively associated with complexity and
density of ornaments. Additionally, angry and high activity performances showed
more pre-main timing than on-main timing. This correlation with timing is different
from that found in the flute performances, which showed a relationship between
timing and valence. The correlations observed for the violin may relate to the higher
density of ornaments for active emotions, because there is a restriction on the
ornaments that can be timed on-main. Therefore, if more notes are added, relatively
many will occur before the original time of the main note (pre-main).
Violin: Sound level of performances. The average relative sound level of a
performance was negatively associated with sad performances, and positively with
angry and high activity performances (see but-last row in Table 2). This indicates that
the violinist used sound level as a significant parameter for signaling intended
emotion, in contrast with the flutist, whose performances did not show any significant
correlations with sound level.
Violin: Structural position of ornaments. As was the case with the flutist, the only
significant correlations between instructed emotion and structural position were with
metrical position (see last row in Table 2). Metrical position was positively associated
with sad performances and, negatively with angry performances and activity. This is
somewhat different from the flutist, who also showed correlations with valence and
happiness. The correlations between metrical position and sad, angry and activity for
22
the violinist may again be related to the larger density of ornaments for the active
emotions: When more short notes are added, more of these notes fall on weak beats.
Flute and Violin: Effect of emotion intensity. In the analyses above, no distinction was
made between mild and intense emotions. To explore the effect of intensity, a
principal component analysis was performed using correlations between the
characteristics of the ornamentation in each of the 24 conditions (4 emotions * 2
intensities * 3 fragments). Only the characteristics that showed a significant effect of
emotion in the earlier analyses were included (see Tables 1 and 2). The two main
components with an eigenvalue greater than 1 were selected for rotation. Figure 5
shows the scores of each performance on the first two rotated factors.
Insert Figure 5 around here
For the flutist, the first two dimensions accounted for 55% of the variance.
Dimension 1 was most strongly correlated with the presence of single appoggiaturas
and negatively correlated with the complexity of ornaments (r = .94 and -.93,
respectively), while dimension 2 of the flutist most strongly correlated with the
presence of slides and pre-main rather than on-main timing (r = .84 and .81,
respectively).
As can be seen in the top panel of Figure 5, the four emotions tended to occupy
different spaces: Sad performances loaded high on dimension 1, and low on
dimension 2. Loving and happy performances scored in the middle of dimension 1
and high (loving) or medium high (happy) on dimension 2. Finally, angry
performances loaded low on both dimensions. The intense and mild versions of a
performance were not always grouped closely together. Overall, there was a tendency
23
for the intense version to score lower on dimension 2 than the mild version, which
means that intense versions tended to have fewer slides and more pre-main timing
than mild versions.
For the violinist, the first two components accounted for 57% of the variance.
Dimension 1 was most strongly negatively correlated with timing (higher scores have
more pre-main than on-main timing) and with the metrical position of ornaments (r =
-.84, and -.82, respectively), while dimension 2 was most strongly negatively
correlated with the presence of single appoggiaturas, and positively with the presence
of turns and complexity (r = -.79, .78, and .78, respectively).
As can be seen in the bottom panel of Figure 5, also for the violinist, the four
emotions tended to occupy different spaces. Sad performances scored lowest on
dimension 1, followed by loving and happy performances. Angry performances
scored highest on dimension 1. The scores on dimension 2 were generally
intermediate with some exceptions. Angry performances scored relatively high on
dimension 2, and sad relatively low, although the loving version of fragment 3 scored
lowest. Overall, there was a tendency for intense versions to score higher on
dimension 2 than mild versions of an emotion, which indicates that intense versions
tended to have more complex ornaments, more turns, and fewer single appoggiaturas.
Additionally, for the negative emotions, intense versions tended to score lower on
dimension 1 than mild versions, which indicates that they had relatively more on-
main timing on heavier beats.
Study 2: Perception of emotion in ornamented melodies
Our second study consisted of a listening experiment. In this experiment, the
communication of emotion from performer to listeners was investigated. We
examined the recognition of the emotions by the listeners as well as the relationship
24
between listeners’ responses and applied ornamentation. Participants were not
Baroque specialists, but had considerable musical training in western classical music.
Participants rated the presence of the four emotions--happiness, love, sadness, and
anger--on a seven-point rating scale. Unipolar scales were used to allow for
ambiguous and relatively open responses (e.g., both happy and sad might be
perceived to be present to some extent in a performance, or none of the emotions
might be perceived to be strongly present). No differentiation was made between mild
or intense emotions, in order to keep the rating task simple. Instead, we expected that
an emotion would be perceived to be more strongly present in performances of
intense emotions than performances of mild emotions.
Method
Material. The variously ornamented performances of fragments 1, 3, and 4 that were
analyzed in Study 1 were used in the listening experiment. The recordings of these
performances were overdubbed with the recordings of the piano accompaniment and
presented as a mixed mono-track file to both ears. Because the solo performances and
the piano performances were recorded in the same hall and with the same recording
set up, the resulting sound-mix was of very high quality and convincingly realistic;
the fact that it was produced by overdubbing was not noticed by the listeners.
Participants. Twenty-four music students participated in the experiment. They were
divided into two groups of twelve – the violin and the flute group. The majority were
second year music undergraduates, but some higher-level students participated as
well. They were paid a small amount for their participation. The average age was 24
for the flute group and 23 for the violin group. The average age of the violin group is
given excluding one elderly undergraduate music student (age 66). The average
number of years of instrumental practice was 10 for the flute group and 14 for the
25
violin group. Participants performed a variety of instruments. The flute group
consisted of one string player, three wind instrument players, four voice majors, two
keyboard players, a harpist and a composer. The violin group consisted of four string
players, four wind instrument players, one voice major, and three keyboard players.
Procedure. The participants were tested individually. Assignment to the flute or
violin group was done in alternation. The participants in the flute group listened to the
flute performances in random order and the participants in the violin group listened to
the violin performances in random order. The participant was seated at an iBook
Macintosh computer and was given instructions to read. Some practice with the task
was provided, after which the real experiment would begin. The practice trials
consisted of one expressive and one neutral performance of each fragment. After the
practice trials, there was some time for questions and feedback. The duration of the
entire experiment was less than an hour.
A user interface with a play button and four vertical seven-point rating scales with
verbal labels of the emotions on top, designed in POCO (Honing, 1990), was used to
play the musical material and record the ratings. The order of the four emotions in the
interface was counter-balanced over participants. The music was presented over Sony
dynamic stereo professional headphones, model MDR-7506, at a comfortable volume.
The instruction to the participants was to listen carefully to the performances and
the ornaments that the performer used, and to judge the extent to which the emotions
happy, loving, sad, and angry were perceived to be present in the performance.
Participants were asked to rate the presence of each emotion independently. They
were free to rate the emotions in any order and could change the rating until they
pressed the ok/save button. They were asked to try to listen to each performance only
26
once, although it was possible to listen to it again if needed by pressing the play
button again.
Results
Rating of the presence of the emotions. Two measures of percent correct are reported
here. The target emotion may be said to be correctly identified by the listener when
that emotion is rated as more strongly present in the target performance than the other
emotions. The rating of the target emotion is than absolutely highest. However, a
target emotion may also be said to be correctly identified when that emotion is rated
to be more strongly present in the target performance than in the other performances
of a fragment. The rating of the target emotion is relatively high for the target
performance.
Table 3 shows the mean percent correct scores for the relatively judged emotions,
per emotion and per fragment, for the flute and the violin performances. The absolute
percent correct scores are given in parentheses.
Insert Table 3 around here
The values in Table 3 should be compared to a chance level of 25%. This is the
chance probability that a target emotion is the highest rated emotion. It can be seen
that happy was not always recognized above chance, but the other emotions generally
were; their relative percentages correct were at least 50% and for some conditions
close to 80% or even 90%. These results do not distinguish between conditions where
the instruction was to express a mild or intense version of the emotions. The
difference in percentage correct was small between the versions, although there was a
tendency for the intense emotions to be better recognized than the mild emotions.
27
The differences between the relative and the absolute percentages highlight the
influence of the fragments. For example, F4 was generally judged to be happy in
character, while F1 and F3 were generally judged to be sad in character. This is
suggested by the absolute percentage correct. Such trends, related to the
characteristics of the fragments, are cancelled out in the relative percentage correct.
It is noteworthy that all relative percentages are higher than the absolute
percentages, except for some loving and sad conditions, which is likely a result of an
interchanging of loving and sad, as we will see next.
Correlations were used to examine the association between perceived emotions and
between perceived and instructed emotions. Data was averaged over participants.
Before averaging, however, the effect of fragment on the emotion rating was
separated from the effect of performance by subtracting the average emotion rating of
a fragment from the emotion rating of a performance. Instructed emotions were coded
as 0 or 1, while ratings of perceived emotions varied between 1 and 7. Tables 4 and 5
show the significant correlations for the flute and violin performances, respectively.
Insert Tables 4 and 5 around here
For both the flutist and the violinist, significant positive correlations existed
between the ratings of love and sadness, and between instructed and perceived anger,
instructed and perceived love, instructed and perceived sadness, and between
instructed sadness and perceived love (for the flute performances) or instructed love
and perceived sadness (for the violin performances). In other words, sadness and love
confounded, and happiness was not reliably communicated. Additionally, anger and
love as well as anger and sadness correlated negatively. In contrast to what might be
28
expected, happiness and sadness correlated negatively only in a few instances (e.g.
instructed sadness and perceived happiness).
Relation between performances and ratings. To explore which of the analyzed
characteristics of the performances may have influenced the listeners’ ratings of
emotions, correlations between analyzed characteristics and average ratings per
performance were calculated. This was done for each characteristic and rated emotion
separately, and separately for the flutist and violinist. As before, the effect of fragment
on the emotion rating was first separated from the effect of performance by
subtracting the average emotion rating of a fragment from the emotion rating of each
performance. This was done for participants individually, before averaging the rating
data.
All analyzed characteristics of the ornaments per performance were taken into
account, which included the different types of ornaments, the ornament
characteristics, the sound level of the performance, and the structural position of the
ornaments. Tables 1 and 2 (right columns) show the correlations that were significant
(p < .05). Generally they confirm the relationships observed in the analyses of the
performances, with some exceptions. For example, some effects that were observed in
the performances were not recognized reliably by the listeners, such as the association
between the presence of slides and loving for the flutist or the presence of trills and
happy for the violinist.
Discussion
In this study, we investigated the communication of emotions in modern
performances of a Handel sonata by means of ornamentation. The recognition of the
performers’ intentions by musically trained listeners was successful on the whole.
Sadness, anger and love were communicated well above chance level. The
29
recognition of happiness was, however, less successful and was above chance level
for some fragments only. In the ratings of the presence of the four emotions, sadness
and love were correlated, and instructed sadness and love were perceived as
expressing both love and sadness. No such correlations between the two negative or
the two positive emotions were observed, nor were the two active emotions confused.
In other words, the performers communicated distinguishable categories of emotions
with the exception of the low-activity emotions studied here – sad and loving.
Despite this observation of communication of discrete emotions, we also observed
certain strong correlations between characteristics of ornaments and the activity of
emotions. Ornament density and complexity were positively related to active
emotions for both performers. Additionally, the violin performances showed a
correlation between activity of emotions and sound level and the flute performances
showed a correlation between activity of emotions and the use of trills, arpeggios, and
short as opposed to long durations. In contrast to what might be expected based on
previous literature (Balkwill & Thompson, 1999) complexity was a factor related to
activity of emotions rather than to the valence of emotions.
Only a few relationships with valence were observed. The flutist used more slides
and timed the ornaments with more anticipation, and more often on weak beats in the
performances of positive emotions than in performances of negative emotions. The
violinist used more trills in positive emotion conditions.
Sometimes specific ornaments were related to specific emotional intentions. For
example, the flutist used slides specifically in the loving conditions and turns in the
happy conditions, while the violinist used trills in the happy conditions and substitutes
(to lower pitch ranges) in sad performances. However, most of the results were in line
with our hypothesis that ornaments can function as acoustic cues for emotions; the
30
characteristics of the ornaments such as density, duration, sound level and timing
mattered rather than the type of ornament or the position of the ornament within the
melody. Notably, most of these characteristics are not unique to ornaments, but could
just as well be characteristics of other melody notes or the musical composition in
general.
Some evidence was found for the relevance of the structural position of the
ornaments. The structural position of the ornaments was examined to include an
aspect of the theories of Meyer (1956) and Langer (1942), which emphasize the
relevance of how music unfolds over time for the emotional characteristic of music.
The only significant effect that was found was a significant variation of the metrical
position of ornaments with instructed and perceived emotion. For both performers,
sadness was associated with ornaments at heavy metrical positions, which
corresponds to a frequent use of on-main appoggiaturas in this condition. Ornaments
at weak metrical positions were, on the other hand, associated with the presence of
happiness and positive emotions for the flutist, and with the presence of anger and
active emotions for the violinist. Apparently, the performers applied weak-beat timing
in distinct ways, possibly related to the tendency of the flutist to time ornaments pre-
main in positive emotion conditions, and the tendency of the violinist to add more
ornaments in high activity emotion conditions.
From the observed relationships between expressed or perceived emotion and
ornamentation, it should not be concluded that there was strong evidence for shared
knowledge between performers and listeners about the expression of emotion through
ornamentation. The means to express the emotions through ornamentation were
different for the two performers and, thus the cues to infer emotions were different for
the two groups of listeners. Expressed emotions could be recognized, although not all
31
varied characteristics of ornaments were effective in conveying the performers’
intentions. The disagreement in means of expression between the performers coupled
with a fairly accurate communication between performer and listeners suggest that
conventions concerning communication of emotion through ornamentation are not
strong (at least nowadays), but that does not preclude communication. It seems that,
although 18th century musicians may have had stronger associations between types of
ornaments and their emotional meanings, modern musicians do not clearly possess
this.
It should be noted that this study was still exploratory and needs corroboration. For
example, in the reported experiments, articulation and timbre may have played a role
in the communication of emotions and may have helped participants to infer emotions
correctly. To assess the contribution of ornamentation more strictly, performers may
be asked, in a follow-up study, to perform the ornamented versions without
knowledge about the intention to express different emotions, where after listeners rate
the perceived emotions. The observed associations between ornamentation and
emotions should therefore be taken as suggestive only.
Moreover, the results of this study relied heavily on only two performers and were
probably to some extent specific to the musical style of the Handel sonata. A larger-
scale study with more performances with fewer ornamentation per fragment may be
needed here, also to examine whether some of the differences between the performers
have greater generality for their respective instruments.
Although the performers were generally comfortable with their task, they did
mention that it was a bit awkward to impose an emotion that contrasted with the mood
of the original music. Specifically, the violinist found it odd to perform the music
32
angrily, when originally there was no anger in the music. Nevertheless, the
communication of instructed anger was most successful.
These and related issues of ecological validity complicate systematic research, but
nevertheless need careful consideration. The use of both systematic variation of
variables and examples from existing music may provide a solution in future research.
Given certain restrictions, our study showed that modern musicians are able to
reliably use ornamentation for the communication of emotion and suggested how this
communication might work. It emphasized performers’ varying strategies to express
emotions, the predominance of communication of activity, and the predominance of
cue-like communication. Even though the strategies of ornamentation may have
limited generality, communication still functioned properly.
References
Bach, C.P.E. (1753/1949). On the true art of playing keyboard instruments.
Translation from German: W.J. Mitchell. London: Cassell.
Balkwill, L.L. & Thompson, W.F. (1999). A cross-cultural investigation of the
perception of emotion in music: Psychophysical and cultural cues. Music
Perception, 17, 43-64.
Boersma, Paul & Weenink, David (2006). Praat: doing phonetics by computer
(Version 4.3) [Computer program]. Retrieved February 2005, from
http://www.praat.org/.
Clynes, M. (1977). Sentics: The touch of emotion. New York: Doubleday Anchor.
Donington, R. (1963). The interpretation of early music. London: Faber and Faber.
Gabrielsson, A. (2001/2002). Perceived emotion and felt emotion: Same or different?
Musicae Scientiae, special issue 2001/2002, 123-147.
33
Gabrielsson, A. & Juslin, P.N. (1996). Emotional expression in music performance:
Between the performer’s intention and the listener’s experience. Psychology of
Music, 24, 68-91.
Gabrielsson, A. & Juslin, P.N. (2003). Emotional expression in music. In Davidson,
R. J., Scherer, K.S., & Goldsmith, H. H. (eds). Handbook of affective sciences
(pp. 503-534). New York: Oxford University Press.
Gabrielsson, A., & Lindström, E. (2001). The influence of musical structure on
emotional expression. In Juslin, P.N., & Sloboda, J.A. (eds), Music and
emotion: theory and research (pp. 223-248). New York: Oxford University
Press.
Gagnon, L., & Peretz, I. (2003). Mode and tempo relative contributions to "happy-
sad" judgements in equitone melodies. Cognition and Emotion, 17, 1-25.
Hevner, K. (1935). Expression in music: A discussion of experimental studies and
theories. Psychological Review, 42, 186-204.
Honing, H. (1990). POCO: an environment for analysing, modifying and generating
expression in music. In Proceedings of the 1990 International Computer
Music Conference (pp.364-368). San Francisco: Computer Music Association.
Ilie, G. & Thompson, W.F. (2006). A comparison of acoustic cues in music and
speech for three dimensions of affect. Music Perception, 23, 319-329.
Juslin, P.N. (2000). Cue utilization in communication of emotion in music
performance: Relating performance to perception. Journal of Experimental
Psychology: Human Perception and Performance, 26, 1797-813.
Juslin, P. N., & Laukka, P. (2001). Impact of intended emotion intensity on cue
utilization and decoding accuracy in vocal expression of emotion. Emotion, 1,
381-412.
34
Juslin, P.N., & Laukka, P. (2003). Communication of emotions in vocal expression
and music performance: Different channels, same code? Psychological
Bulletin, 129 (5), 770-814.
Kramer, E. (1964). Elimination of verbal cues in judgments of emotion from voice.
Journal of Abnormal and Social Psychology, 68, 390-396.
Krumhansl, C.L. (1996). A perceptual analysis of Mozart’s Piano Sonata K.282 –
Segmentation, tension and musical ideas. Music Perception, 13, 401-432.
Krumhansl, C.L. (1998). Topic in music: An empirical study of memorability,
openness, and emotion in Mozart’s String Quintet in C Major and Beethoven’s
String Quartet in A Minor. Music Perception, 16. 119-134.
Lerdahl, F. & Krumhansl, C.L. (2007). Modeling tonal tension. Music Perception, 24,
329-366.
Langer, S.K. (1963/1942). Philosophy in a new key. A study in the symbolism of
reason, rite and art. Cambridge, Massachusetts: Harvard University Press. 3rd
Edition.
Mandler, G. (1975). Mind and emotion. New York: Wiley.
Mattheson, (1739/1981). Der vollkommene Capellmeister. Translation into English:
E.C. Harriss. Ann Arbor, Michigan: UMI Research Press.
Meyer, L.B. (1956). Emotion and meaning in music. Chicago: The University of
Chicago Press.
Oatley, K. (1992). Best-laid schemes: The psychology of emotions. Cambridge:
Cambridge University Press.
Ortony, A., Clore, G.L. & Collins, A. (1988). The cognitive structure of emotions.
Cambridge: Cambridge University Press.
35
Quantz, J.J. (1752/1966). On playing the flute. Translation from German: E.R. Reilly.
London: Faber.
Russell, J.A. (1980). A circumplex model of affect. Journal of Personality and Social
Psychology, 39, 1161-1178.
Scherer, K.R., Banse, R., Wallbott, H.G., & Goldbeck, T. (1991). Vocal cues in
emotion encoding and decoding. Motivation and Emotion, 15, 123-148.
Scherer, K.R. & Zentner, M.R. (2001). Emotional effects of music: production rules.
In Juslin, P.N. & Sloboda, J.A. (eds), Music and emotion: theory and research
(pp. 361-392). New York: Oxford University Press.
Scherer, K.R. (2004). Which emotions can be induced by music? What are the
underlying mechanisms? And how can we measure them? Journal of New
Music Research, 33, 239-251.
Schubert, E. (2001). Continuous measurement of self-reported emotional response. In
Juslin, P.N. & Sloboda, J.A. (eds), Music and emotion: theory and research
(pp. 393-414). New York: Oxford University Press.
Sloboda, J.A., & Juslin, P.N. (2001). Psychological perspectives on music and
emotion. In Juslin, P.N. & Sloboda, J.A. (eds), Music and emotion: theory and
research (pp. 71-104). New York: Oxford University Press.
Sloboda, J.A., & Lehmann, A.C. (2001). Tracking performance correlates of changes
of perceived intensity of emotion during different interpretations of a Chopin
piano prelude. Music Perception, 19, 87-120.
Sundberg, J., Iwarsson, J. & Hagegård, H. (1995). A singer’s expression of emotions
in sung performances. In Fujimura, O. (ed), Vocal fold physiology: Voice
quality control (pp. 217-231). San Diego: Singular Publishing.
36
Trehub, S.E. & Nakata, T. (2001/2002). Emotion and music in infancy. Musicae
Scientiae, special issue 2001/2002, 37-61.
Trainor, L.J. & Heinmiller B.M. (1998). The development of evaluative responses to
music: Infants prefer to listen to consonance over dissonance. Infant Behavior
& Development, 21, 77-88.
Wessel, F.T. (1979). Affectenlehre in the 18th century. Ann Arbor, Michigan:
University Microfilms International.
Author notes
Renee Timmers, School of Music, Northwestern University; Richard Ashley,
School of Music, Northwestern University.
Renee Timmers is now at Music, Mind, Machine group, Nijmegen Institute for
Cognition and Information, Radboud University Nijmegen, Nijmegen, The
Netherlands.
This research was supported by a Talent stipend from the Netherlands Scientific
Organization. We wish to thank the performers for their excellent contributions, and
the participants for their willingness. We thank Zohar Eitan, Henkjan Honing, Daniel
Leech-Wilkinson and the music departments of Northwestern University and King’s
College for providing the right input and context for the study, and Nicola Dibben,
Bruno Repp and an anonymous reviewer for their excellent suggestions.
Correspondence concerning this article should be addressed to Renee Timmers,
Nijmegen Institute for Cognition and Information, Radboud University Nijmegen,
P.O. Box 9104, 6500 HE Nijmegen, The Netherlands. E-mail: [email protected] or
37
Footnotes
Footnote 1. CD: George Frideric Handel (1685-1759) Complete Wind Sonatas,
Philips Classics Production 446 563-2 (1995), Printed and made in U.S.A. Original
recordings 1/1981 (CD1 track 9&11), 5/1983 (CD2). Performed by Academy of St
Martin in the Fields: F major (HWV363a) and C minor (HWV366) – Neil Black
Oboe, George Malcolm harpsichord, Graham Sheen bassoon; C major (HWV365) –
Michala Petri recorder, George Malcolm harpsichord, Denis Vigay cello; B minor
(HWV376) – William Bennett flute, Nicholas Kremer harpsichord, Denis Vigay cello.
Footnote 2. According to Bach, appoggiaturas always take something from the
duration of the main note, although this might be very little if the ornament is short.
Footnote 3. A web-link to PRAAT and explanatory texts can be found at
www.fon.hum.uva.nl/praat/.
38
Table 1: Correlations for flute performances between emotions (Columns) and
ornamentation (Rows). Significant correlations are in bold (N = 24, p < .05).
Emotion* i-H i-L i-S i-A i-Act i-Val r-H r-L r-S r-A
Single -.09 -.02 .59 -.48 -.49 -.09 -.42 .53 .70 -.62
Double -.27 -.03 .11 .18 -.07 -.26 .13 -.01 .07 .03
Trill .17 .13 -.60 .30 .41 .26 .30 -.44 -.66 .54
Turn .41 -.35 -.10 .05 .39 .05 .10 -.19 -.16 .12
Mordent .05 -.23 -.12 .29 .30 -.15 -.05 -.31 -.21 .39
Slide .00 .55 -.17 -.38 -.33 .47 .14 .32 -.05 -.17
Arpeggio .08 -.11 -.42 .45 .46 -.03 .22 -.50 -.55 .41
Substitute .08 .04 -.28 .17 .21 .10 .08 -.26 -.19 .04
Complexity .07 .01 -.55 .46 .46 .08 .41 -.49 -.68 .62
Density -.25 .42 -.27 .10 -.13 .15 .56 .08 -.27 -.05
On-timing -.41 -.28 .48 .21 -.17 -.60 -.22 -.07 .35 .09
Duration -.32 .17 .61 -.46 -.67 -.13 -.25 .57 .64 -.48
Direction .18 .25 .02 -.45 -.24 .37 .00 .55 .25 -.59
Harmony -.02 .20 -.55 .38 .31 .15 -.36 -.45 -.56 .52
Sound level .02 .17 -.10 -.10 -.06 .17 .11 .06 -.13 .18
Meter -.42 -.32 .79 -.05 -.41 -.64 -.47 -.20 .62 .21
Form pos -.03 -.17 .41 -.22 -.21 -.17 -.19 .18 .33 -.14
Bar pos .28 -.29 .04 -.04 .22 .00 .03 -.11 -.17 .05
* Emotions are instructed (i-H, i-L, i-S, i-A, i-Act, i-Val) or rated (r-H, r-L, r-S, r-A).
39
Table 2: Correlations for violin performances between emotions (Columns) and
ornamentation (Rows). Significant correlations are in bold (N = 24, p < .05).
Emotion* i-H i-L i-S i-A i-Act i-Val r-H r-L r-S r-A
Single app. -.28 .43 .11 -.26 -.47 .13 -.39 .18 .26 -.14
Double app. .39 -.23 -.58 .43 .70 .13 .23 -.37 -.48 .36
Trill .64 -.11 -.29 -.25 .34 .47 .34 .12 -.08 -.16
Turn -.19 .11 -.19 .27 .07 -.07 .53 -.07 -.27 .22
Mordent .01 -.09 -.11 .19 .17 -.08 .35 -.06 -.23 .14
Slide .19 -.19 -.19 .19 .33 .00 .14 -.18 -.18 .24
Arpeggio .08 -.13 .03 .02 .09 -.05 -.15 -.04 -.19 .02
Substitute -.24 -.08 .55 -.23 -.40 -.28 -.27 .16 .48 -.29
Complexity .19 -.26 -.47 .54 .63 -.06 .57 -.34 -.53 .44
Density .12 -.04 -.53 .45 .49 .07 .16 -.42 -.67 .54
On-timing .04 .19 .32 -.55 -.44 .20 -.10 .39 .46 -.41
Duration .11 -.26 .21 -.06 .04 -.13 .05 -.08 .09 -.02
Direction -.08 -.02 -.03 .13 .04 -.09 .12 -.09 .08 .04
Harmony .03 .28 -.17 -.13 -.09 .27 .04 .08 -.01 .09
Sound level .03 .02 -.76 .71 .64 .04 .42 -.61 -.70 .74
Meter -.09 .23 .46 -.60 -.60 .12 -.14 .42 .51 -.47
Form pos .45 -.18 -.25 -.02 .37 .24 .27 -.02 .04 -.10
Bar pos -.10 -.27 .29 .08 -.03 -.32 .07 -.11 .01 .02
* Emotions are instructed (i-H, i-L, i-S, i-A, i-Act, i-Val) or rated (r-H, r-L, r-S, r-A).
40
Table 3: Average percentage of participants that gave the relatively highest (or in
parentheses the absolutely highest) rating to the target emotion.
Flute Violin
F1 F3 F4 F1 F3 F4
Happy 21 (17) 38 (8) 46 (54) 42 (8) 21 (4) 63 (46)
Loving 50 (42) 54 (58) 75 (58) 50 (58) 54 (46) 50 (21)
Sad 67 (75) 88 (79) 58 (21) 50 (79) 71 (88) 54 (50)
Angry 63 (50) 58 (33) 54 (50) 92 (67) 92 (50) 79 (83)
41
Table 4: Correlations between rated emotions (rows) and rated or instructed
emotions (columns) for the flute performances. Significant correlations are in bold (N
= 24, p < .05).
Rated Instructed
Happy Loving Sad Happy Loving Sad Angry
Happy .32 .30 -.59 -.04
Loving .06 -.13 .50 .44 -.80
Sad -.48 .69 -.27 .09 .78 -.60
Angry -.03 -.89 -.70 .01 -.39 -.42 .79
42
Table 5: Correlations between rated emotions (rows) and rated or instructed
emotions (columns) for the violin performances. Significant correlations are in bold
(N = 24, p < .05).
Rated Instructed
Happy Loving Sad Happy Loving Sad Angry
Happy .34 .15 -.47 -.02
Loving .26 .20 .41 .26 -.88
Sad -.14 .70 .02 .70 .55 -.75
Angry -.07 -.93 -.83 -.27 -.23 -.40 .91
43
Appendix 1: Track information of CD sent to the performers
Track Handel sonata, movement Comments
1 F-major (HWV363a), mv5 happy, mild (calm joy, gayement)
2 F-major (HWV363a), mv4 happy, intense (euphoric, brillant)
3 F-major (HWV 363a), mv1 loving, mild (tender, cantabile)
4 C-major (HWV 365), mv1 loving, intense (passionate, con brio)
5 B minor (HWV 376), mv3 sad, mild (regret, doloroso)
6 C minor (HWV 366), mv3 sad, intense (distress, lagrimoso)
7 B minor (HWV 376), mv1 anger, mild (boldness, ardito)
8 C minor (HWV 366), mv2 anger, intense (rage, furioso)
9 G-minor (HWV 360), mv1 piano performance of basso continuo
10 G-minor (HWV 360), mv2 piano performance of basso continuo
11 G-minor (HWV 360), mv3 piano performance of basso continuo
12 G-minor (HWV 360), mv4 piano performance of basso continuo
44
Figure captions
Figure 1: Scores of fragments 1, 3 and 4. The last chord of fragment 3 is omitted to
save an extra line.
Figure 2: Examples of the different types of ornaments distinguished.
Figure 3: Examples of the ornamentation used in this study. All examples show the
first two bars of F1 and are taken from the intense version of an emotion.
Figure 4: Schematics of analysis procedure of the ornamentation added by the
performers to the original music of the Handel sonata.
Figure 5: Multiple dimensional scaling plot of the flute performances (top) and violin
performances (bottom). Instructed emotions are indicated by initials. Capital initials
indicate intense emotions and lower case initials indicate mild emotions. Lines
connect mild and intense versions of a performance. Different line types indicate the
different fragments.
45
Fragment 1
Fragment 3
Fragment 4
Figure 1
46
Figure 2
47
Figure 3
48
Figure 4
49
Figure 5